Electronic Design

Bob's Mailbox

I read your article “What’s All This Analog Engineering Stuff, Anyhow?” (Oct. 2, p. 18, ED Online 19754) I totally agree that the need for trained analog engineers is not going away. (I am not so interested in training, but in education. /rap)

I have been in analog engineering since my BSEE in 1958 from that other school in Pasadena, and I can still analyze a circuit using the math tools I got in school. (I don’t use “PSPICE” or any other kind of hired computer analysis. I use “Pease SPICE,” which is “back of envelope SPICE.” I use analog insights to analyze how the response will be. /rap)

The problem is, when I interview candidates for an engineering position, I find that many cannot even draw the output waveform on a simple RC differentiator driven by a step function. If I ask them if they know how to use the Laplace transform, they say they haven’t looked at it since school.

(Actually, I don’t think anybody ever tried to teach me Laplace transform stuff. But I do know how to close a loop. Did you see my BOBB, Ball On Beam Balancer? Hey, a ball on a beam is a double integrator. Most people don’t know how to close a loop around that. /rap)

A frightening majority of the candidates I see are now relying on the computer and simulation software to do their designs. The trend is worrisome. –BRUCE WILKINSON

There are many trends that are worrisome. We have to encourage the smart kids that do know how to solve problems and comprehend and analyze circuits. We don’t need a lot of well-trained engineers, just a few well-educated guys. –RAP

I read your book Troubleshooting Analog Circuits, and something popped up in my mind. On page 101-102, you wrote about manipulating noise gain to get stability. You put the resistor from negative input to ground or to positive input. The result was noise gain increasing to 6×, but signal gain is still 1×. I’m interested in noise gain, say, the whole system is already stable. (The system may be stable. But if the noise gain is increased a lot, the output can get slow and/or noisy. /rap) In your examples, the noise gain is always greater than signal gain. (That is often true. But, no, for a unity gain follower, the NG can be equal to the signal gain. Or for an amplifier with positive gain, such as +1, +2, +10, or +100, the NG can be equal to or greater than the signal gain. It does not have to be greater than... /rap) Is there an op-amp input/feedback resistor configuration that can give noise gain that’s less than signal gain? –DAVID

Normally, the answer is no. If you make a logger using a transistor with a grounded base, with the op amp’s output coupled in to the emitter, and the collector goes to the summing point, the transistor can add gain to the loop. The noise gain can stay low, while the gain gets higher. But that happens only because the transistor is adding extra gain. Sort of cheating. So unless you are making a log function, NG is normally equal to or higher than the signal gain—and can be a lot higher. –RAP

We have a battery-powered LED driver circuit that will be part of a disposable medical device. The logical question came up about whether leakage through the transistor switch would drain the battery in two years of product shelf life. If I can believe it, PSPICE shows an off-state leakage of 39 nA, not much at all (Most transistors, when OFF, leak less than 39 pA. But PSPICE is a poor choice to trust. /rap), and if true with the actual circuit, two-year shelf life would be a piece of cake.

Figuring that the real world has lots of potential leakage paths on a printed-circuit board (PCB), I thought it best to make an actual measurement. (Any good PCB will also leak less than 39 pA. /rap) The equipment at hand is a Fluke 8846A, which touts nanoampere measurement capability. (It will take me a while to find the 8846A. Why not put 1 µF of mylar across the inputs of the Fluker? /rap)

All’s great, except for one issue: noise. Even after zeroing the meter, it’s hard to tell if I am really measuring anything but noise on the 100-µA FS range. (If you really want to read leakage current, use the 1-M or 10-M input in the voltage mode. You can resolve 100 pA × 1 M = 100 µV. You could put 0.1 µF across that to cut the noise down. /rap)

We thought about putting the circuit in a Faraday cage but haven’t done that yet. Do you have a suggestion on making this measurement? –ART ZIKORUS

Put it in a cake pan. A metal cake pan. Put aluminum foil over the top. If you only have a glass cake pan, put tinfoil over the top and the bottom. Ground the tin foil. Any good DVM will be able to resolve sub-nano-ampere currents. –RAP

Comments invited! [email protected] —or: Mail Stop D2597A, National Semiconductor P.O. Box 58090, Santa Clara, CA 95052-8090

BOB PEASE obtained a BSEE from MIT in 1961 and is Staff Scientist at National Semiconductor Corp., Santa Clara, Calif.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.